Journal of Psychiatric Research 38 (2004) 37–46
www.elsevier.com/locate/jpsychires
Electrophysiological evidence of corollary discharge dysfunction in
schizophrenia during talking and thinking
Judith M. Forda,b,*, Daniel H. Mathalonc,d
a
Department of Psychiatry & Behavioral Sciences Stanford University School of Medicine, Stanford, CA 94305-5550, USA
b
Psychiatry Service, Veterans Affairs Palo Alto Health Care System, Palo Alto, California, USA
c
Department of Psychiatry, Yale University School of Medicine, New Haven, Connecticut, USA
d
Psychiatry Service, Veterans Affairs West Haven Health Care System, West Haven, Connecticut, USA
Received 25 March 2002; received in revised form 27 September 2002; accepted 25 October 2002
Abstract
Failure of corollary discharge, a mechanism for distinguishing self-generated from externally-generated percepts, has been posited to underlie certain positive symptoms of schizophrenia, including auditory hallucinations. Although originally described in the
visual system, corollary discharge may exist in the auditory system, whereby signals from motor speech commands prepare auditory
cortex for self-generated speech. While associated with sensorimotor systems, it might also apply to inner speech or thought,
regarded as our most complex motor act. We had four aims in the studies summarized in this paper: (1) to demonstrate the corollary
discharge phenomenon during talking and inner speech in human volunteers using event-related brain potentials (ERPs), (2) to
demonstrate that the corollary discharge is abnormal in patients with schizophrenia, (3) to demonstrate the role of frontal speech
areas in the corollary discharge during talking, and (4) to relate the dysfunction of the corollary discharge in schizophrenia to
auditory hallucinations. Using EEG and ERP measures, we addressed each aim in patients with schizophrenia (DSM IV) and
healthy control subjects. The N1 component of the ERP reflected dampening of auditory cortex responsivity during talking and
inner speech in control subjects but not in patients. EEG measures of coherence indicated inter-dependence of activity in the frontal
speech production and temporal speech reception areas during talking in control subjects, but not in patients, especially those who
hallucinated. These data suggest that a corollary discharge from frontal areas where thoughts are generated fails to alert auditory
cortex that they are self-generated, leading to the misattribution of inner speech to external sources and producing the experience of
auditory hallucinations.
# 2003 Elsevier Ltd. All rights reserved.
Keywords: Schizophrenia; Corollary discharge; N1; EEG coherence
1. Introduction
Self-monitoring is a fundamental element of normal
cognitive and motor functioning. It allows us to do online modifications and corrections of actions. Von Holst
and Mittelstadt (1950) and Sperry (1950) suggested that
motor actions are accompanied by a ‘‘corollary discharge’’ to sensory cortex, signaling that impending
sensations are self-initiated or self-generated. In the
visual system, a corollary discharge may serve to stabilize the visual image during eye movements, maintaining
visuo-spatial constancy. In the somatosensory system, it
* Corresponding author. Tel.: +1-650-493-5000x65249, Fax: +1650-493-4901.
E-mail address: [email protected] (J.M. Ford).
0022-3956/$ - see front matter # 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0022-3956(03)00095-5
may explain why we cannot tickle ourselves (Blakemore
et al., 1998). It has been suggested that the corollary
discharge is an efference copy of a planned action sent
through a ‘‘feed forward’’ mechanism to the appropriate sensory cortex, preparing it for the arrival of the
sensation. In its simplest form, the efference copy works to
suppress perception when it results from a self-generated
action. Thus, in addition to serving as a mechanism for
learning and fine-tuning our actions, the efference copy
may allow an automatic distinction between internallyand externally-generated percepts.
A similar mechanism may exist in the auditory system:
corollary discharges from motor speech commands prepare
auditory cortex for self-generated speech, linking regions
of the frontal lobes where speech is generated to regions
of the temporal lobes where it is heard (Creutzfeldt et al.,
38
J.M. Ford, D.H. Mathalon / Journal of Psychiatric Research 38 (2004) 37–46
1989b; Paus et al., 1996). Creutzfeldt et al. (1989b) had
patients talk and listen to others talking during a presurgical planning procedure while recording from the
exposed surface of the right and left temporal cortices.
Different populations of neurons in both the superior
(STG) and middle temporal gyri (MTG) responded to
speech when it was generated by the subject than when
it was spoken by others. Muller-Preuss and Ploog (1981)
recorded from monkeys and also described differential
responses of STG to self- and other-generated vocalizations. Of particular interest was the suppression of ongoing cortical activity during self vocalization. About
one third of MTG neurons and some STG neurons
showed reduced responsiveness to self-produced speech
in the Creutzfeldt et al study. In the Muller-Preuss and
Ploog study, more than half of the STG neurons were
reduced in responsiveness during vocalization. Although
corollary discharge is typically associated with sensorimotor systems, an association with thinking is plausible, to
the extent that thinking can be regarded as ‘‘our most complex motor act’’ (Jackson, 1958). Indeed, it has been postulated (see page 196, Feinberg & Guazzelli, 1999) that
thinking ‘‘might conserve and utilize the computational and
integrative mechanisms evolved for physical movement.’’
In Fig. 1 we illustrate the concept of the corollary discharge.
These direct studies of cortical activation and deactivation in monkey and human are consistent with more
recent hemodynamic brain imaging studies of word
generation. In the word generation condition, subjects
are presented with initial letters of words and must
generate an exemplar. Compared to simple repetition of
a word, generating a word results in relatively more
activation of frontal lobe structures and relatively less
activation (relative ‘‘deactivation’’) of temporal lobe
structures (Frith et al., 1991, Warburton et al., 1996).
However, during the experience of hearing voices when
none are spoken (auditory verbal hallucinations), temporal
lobe structures are activated, not deactivated in schizophrenics (Dierks et al., 1999, Shergill et al., 2000). This
is indirect evidence that the corollary discharge from the
frontal speech areas is not working to inform the temporal lobes that the input is self-generated.
It has been suggested that failures of this mechanism
may contribute to the positive symptoms of schizophrenia (Feinberg, 1978, Feinberg & Guazzelli, 1999).
Specifically, if an efference copy of an intended action
(or thought) is not sent to the appropriate sensory cortex,
patients may fail to distinguish between their own and
externally generated actions or thoughts, resulting in
passivity experiences or auditory verbal hallucinations.
In this report we summarize the results of a series of
studies using electrophysiological techniques to probe the
brain during self-generated and inner-speech. Three of the
studies used the auditory N1 to assess auditory cortical
responsiveness. The auditory N1, and its magnetic
counterpart the N1m, is usually followed by another
major response, the P2, which has not been extensively
investigated, and will not be discussed here. N1 is generated
in auditory cortex by transient auditory stimuli and it
reaches its peak approximately 100 ms after stimulus onset.
N1 has a long ( 10 s) temporal recovery function, with
smaller N1s associated with shorter inter-stimulus intervals
(ISI) (Davis & Zerlin, 1966). Thus, N1 to an acoustic
probe will be sensitive to the presence of other competing auditory events, or acoustic interference. Addition-
Fig. 1. Schematic showing the hypothesized corollary discharge
mechanism operating during talking. It is shown going from frontal
lobes, where speech is generated, to auditory cortex in the temporal
lobe, where speech is perceived. On the top, the corollary discharge is
shown functioning normally in a control subject; the frontal lobe is
shown in red to indicate activity during talking and auditory cortex is
shown in blue to indicate suppression of activity during talking. On
the bottom, it is shown as functioning abnormally in a patient with
schizophrenia, where activity of auditory cortex is not suppressed
during talking.
J.M. Ford, D.H. Mathalon / Journal of Psychiatric Research 38 (2004) 37–46
ally, N1 is sensitive to attention, being smaller when attention is directed away from the eliciting stimulus (Hillyard et
al., 1973). Importantly, N1 to tone probes presented while
the subject was instructed to listen to pre-recorded speech
was reduced compared to N1s elicited during instructions
to ignore the speech (Papanicolaou et al., 1988). Like the
auditory N1, the visual N1 emanates from sensory cortical
structures (Martinez et al., 1999) and is also affected by
attention, being smaller when attention is directed away
from the eliciting stimulus (Han et al., 2000).
Recent magnetoencephalographic (MEG) studies
measuring N1m have shown that while a subject is
talking, responsiveness of the auditory cortex to 1000
Hz tone probes is dampened and delayed compared to
while a subject is reading silently (Numminen et al.
1999). To rule out the effects of acoustic interference on
responsiveness to probes presented while the subject was
talking aloud, one study (Curio et al., 2000) assessed
responsiveness to vowel sounds as they are being spoken
compared to when they were being played back, a technique
used by us in Experiment 2. Curio et al. found that responses are dampened and delayed during talking compared to
during playback, and they attributed the reduction during
talking to the dampening effect of the corollary discharge of
the planned speech transmitted from pre-frontal speech
areas to temporal lobe auditory processing areas.
We had four aims in these studies: (1) to demonstrate
the corollary discharge phenomenon during talking and
inner speech in human volunteers using event-related
brain potentials (ERPs), (2) to demonstrate that the
corollary discharge is abnormal in patients with schizophrenia, (3) to demonstrate the role of frontal speech
areas in the corollary discharge during talking, and (4)
to relate the dysfunction of the corollary discharge in
schizophrenia to auditory hallucinations.
39
in neurological sequelae) or neurological or other medical illnesses compromising the central nervous system.
Patients were recruited from community mental
health centers, as well as from inpatient and outpatient
services of the Palo Alto Veterans Affairs Health Care
System. Controls were recruited by newspaper advertisements and word-of-mouth, and screened by telephone using the psychiatric screening questions from
the Structured Clinical Interview for DSM-IV.
Patient symptoms were assessed by at least two trained
raters (including a psychiatrist or clinical psychologist)
administering the 18 item Brief Psychiatric Rating Scale
(BPRS) (Hedlund & Vieweg, 1980, Overall et al., 1967).
This was done during a semi-structured interview conducted typically on the same day or within the same
week of ERP testing. Ratings were averaged over two
raters. In addition, the Schedule for Assessment of
Positive Symptoms (SAPS) (Andreasen, 1984) was
administered in the same rating session as the BPRS.
2.2. ERP procedure
2.2.1. ERP Recording
Electroencephalogram (EEG) was recorded from
various scalp sites; but only ERPs recorded from the
subset of scalp sites where auditory ERPs are typically
largest, are presented here. Vertical electro-oculogram
(VEOG) was recorded from electrodes placed above
and below the right eye, and horizontal (HEOG) from
electrodes placed at the outer canthus of each eye. EEG
and EOG were sampled every 2 ms. During acquisition,
EEG data were band pass filtered between 0.05 and 40
Hz. The EEG signal elicited by the probe stimulus was
processed using a variety of techniques to enhance signal to noise ratio and minimize noise due to artifacts
associated with eye movements and speech. Technical
details are available in the original reports.
2. General methods
2.1. Subjects
3. Experiment 1: responses to probes during talking
Table 1 describes the patients and controls participating in each study. Medicated patients with schizophrenia (DSM-IV (SCID1) (First et al., 1995) and
healthy adult comparison subjects (SCID screened for
any significant history of Axis I psychiatric illness)
participated. All gave written informed consent after
procedures had been fully explained. Prospective patient
and control participants were excluded if they met
DSM-IV criteria for alcohol or drug abuse within 30
days prior to study. In addition, patient and control
participants were excluded for significant head injury
(loss of consciousness greater than 30 min or resulting
3.1. Introduction
1
In a few cases, a psychiatrist made the diagnosis by patient chart
review.
Subjects participating in this experiment are described
in Table 1.
The purpose of this study was to compare auditory
cortical responsiveness to probes presented while subjects sat silently (Baseline), spoke out loud (Talking),
and while they heard their speech played back to them
(Listening). We predicted that responses to acoustic
probes would be dampened during talking in control
subjects, but not in patients. The details of this study
appear in our earlier report (Ford et al., 2001c).
3.2. Procedure
40
J.M. Ford, D.H. Mathalon / Journal of Psychiatric Research 38 (2004) 37–46
Table 1
Subject characteristics
Control subjects
Gender (M/F)
Age (years)
BPRS Total
Medication (Conventional/Novel)
Schizophrenic patients
Studies 1 & 4
(n=10)
Study 2
(n=7)
Study 3
(n=15)
Studies 1 & 4
(n=12)
Study 2
(n=7)
Study 3
(n=15)
Mean
(range)
Mean
(range)
Mean
(range)
Mean
(range)
Mean
(range)
Mean
(range)
9/1
44.5 (30–52)
–
7/0
35.9 (26–56)
–
13/2
44.7 (20–58)
–
11/1
39.5 (24–53)
38.5 (21–52.5)
4/8
7/0
34.1 (23–52)
42.6 (19–61)
0/7
13/2
40.1 (26–56)
39.5 (21–52.5)
4/11
Three equiprobable probes, each 250 m in duration,
were presented at random ISIs (0.8, 1.0 or 1.2 s): speech
syllable [ba], broadband noise, and square checkerboard. During the Baseline Condition, the intensity of
[ba] and noise was set to 76 dB SPL (C scale). During
Talking and Listening, the probe loudness was adjusted
upwards to ensure probe discriminability.
3.2.1. During the baseline condition
Subjects sat upright in comfortable chair in a sound
attenuated room, wore foam ear-cuff headphones to
hear the sounds and faced a video monitor to see the
checkerboard. They were asked to keep their eyes
focused on a fixation point on the screen throughout the
sequence of sounds and checkerboards. The presentation of stimuli lasted 2 min and 42 s.
3.2.2. During listening and talking conditions
The same sequence of probe stimuli was presented
while subjects alternated between listening to their own
prerecorded voice repeating a hallucinatory statement
(e.g., ‘‘Get off your duff and do something’’) for 30 s
followed by repeating that same statement for another
30 s. This alternation between listening and talking was
repeated for seven different hallucinatory statements.
Care was taken to ensure comparable intensity of
recorded and spoken sentences, and subjects were
trained to match their speech intensity to that of their
recorded voice. This Listen/Talk sequence lasted a total
of seven min. Although there was a negative emotional
valence to some of the statements, patients were told
before recording that none of the statements was about
them, and after the session, no patients reported being
troubled by the statements.
3.3. Results
The results most relevant to this review report are the
differences between Talking and Baseline. Other results
are reported in more detail elsewhere (Ford et al.,
2001c).
Compared to Baseline, N1 was not reduced during
Talking in patients, but was in controls [GroupCondition: F(1,20)=8.21, P < 0.01]. This can be seen
in Fig. 2. We parsed this interaction and found that N1
was smaller during Talking than during Baseline in the
controls [F(1,9)=23.33, P < 0.001], but not in the
patients [F(1,11)=1.36, P< 0.27]. A GroupCondition
ANOVA for N1 to the checkerboard revealed no significant effects for Group [F(1,20)=0.44, P=0.52],
Condition [F(1,20)=1.63, P=0.22] or the Group
Condition interaction [F(1,20)=0.14, P=0.72].
Correlations of SAPS and BPRS symptom scores with
N1 amplitude during Baseline, Talking, and Talking
minus Baseline were computed. None was significant.
3.4. Discussion
Talking affected N1 to acoustic but not to visual
probes, reflecting the modality specificity of the N1
effects. Furthermore, the pattern of responses to acoustic
probes during Talking and Baseline differed between
controls and patients. In controls, N1 to acoustic probes
was reduced during Talking compared to Baseline. In
patients, N1 to acoustic probes was not smaller during
Talking compared to Baseline. In addition, Baseline N1
amplitude was smaller in patients than in controls.
One of the principal motivations of this study was to
examine the effects of corollary discharge which
according to existing theory should be activated by
talking (Curio et al., 2000) more in controls than in
schizophrenic patients (Feinberg, 1978; Feinberg &
Guazzelli, 1999; Frith & Done, 1989). By its nature, the
corollary discharge happens automatically, without
motivation or effort. However, our measure of cortical
responsiveness, the N1 component, is minimally affected
by attention, and these attention effects on N1 could
possibly obscure our ability to detect the action of the
corollary discharge. That is, N1 to the probe could have
been reduced during talking in the controls because they
found their own voices more interesting than the probe,
while the opposite could have occurred in the patients.
J.M. Ford, D.H. Mathalon / Journal of Psychiatric Research 38 (2004) 37–46
41
Fig. 2. Experiment 1. ERPs to an acoustic probe stimulus (noise) recorded at Cza (mid-way between Fz and Cz) are shown for controls and patients
hearing the probes during a baseline condition of silence and while talking aloud. N1, generated in auditory cortical areas of the temporal lobe, is
denoted with an arrow. The red lines depict auditory cortical activity to the probes during the silent baseline, and the blue lines depict auditory
cortical activity to the probes during talking. In the controls, auditory cortical activity is relatively suppressed (blue) during talking compared to the
silent baseline (red); this is not true in the patients. ERPs were filtered with a 0.5–15 Hz band-pass filter.
4. Experiment 2: responses to talking
4.1. Introduction
Although the effects of attention are minimal on N1,
it was important to rule out the possibility of differential
attention effects to probes and talking. To allow a more
direct assessment of the corollary discharge during talking,
we conducted the next experiment in which talking is the
probe. That is, we assessed brain responses to talking
directly, by eliciting ERPs to the speech sound as it was
being produced, a procedure used by Curio and colleagues
(2000). This procedure can be seen in Fig. 3. Details of this
study appear in an earlier report (Ford et al., 2001a).
4.2. Procedure
See Table 1 for description of subjects included in this
study.
Subjects uttered syllables [a] and [ei] for about 3 min,
after they were instructed about how loud (comfortable
speaking level) and how often (syllable frequency=1/1.5 s,
probability of [a]=0.80) to say the syllables. Each subject’s
self-generated vowel sequence was recorded and played
back to them through head phones, after first adjusting the
gain to equalize loudnesses during playback and talking.
On average, patients uttered 121, and comparison subjects,
118 vowels. EEG epochs of 1 s were synchronized to
speech onset, eye-blink corrected, and further screened
to exclude speech-related artifacts during speaking.
After artifact rejection, about half the remaining trials
were included in the ERP averages. ERPs were collapsed
across [a] and [ei] and filtered (2–8 Hz) to reduce speechrelated artifacts that might affect our measurement of
Fig. 3. Schematic illustrating the paradigm in which we compare
auditory cortical responses to speech sounds as they are being produced, during talking and during playback of the same sounds.
42
J.M. Ford, D.H. Mathalon / Journal of Psychiatric Research 38 (2004) 37–46
N1. N1 amplitude was measured as the maximum
negativity between 40 and 180 ms.
4.3. Results
Effects of speaking and listening on N1 amplitude to
speech sounds differed in patients and comparison subjects [GroupConditionElectrode Site: F(2,24)=5.69,
P < 0.02, two-tailed], with the GroupCondition interaction only being significant at Cz [F(1,12)=4.83,
P < 0.05, two-tailed]. This interaction was due to N1 to
the vowels being smaller as they are being spoken than
when they were played back in the control subjects
[paired t(6)= 2.04, P=0.04, one-tailed], but not in the
patients [paired t(6)=0.84, P=0.22, one-tailed]. In the
patients, N1 during talking was not smaller than during
playback, as can be seen in Fig. 4.
4.4. Discussion
In this study, the normal subjects produced smaller N1s
to uttered than played back vowels. This is consistent with
previous MEG findings (Curio et al., 2000) and provides
neurophysiological evidence in support of a speech-related
corollary discharge suppressing responsiveness of auditory
cortex to self-generated speech sounds. Patients did not
show this reduction in N1 to their own utterances, suggesting that this mechanism of auditory cortex suppression is dysfunctional in schizophrenia.
5. Experiment 3: responses to probes during inner speech
5.1. Introduction
While experiments on talking have the distinct advantage of being verifiable, our interest in the dysfunctional
corollary discharge in schizophrenia is related to the
experience of auditory verbal hallucinations, which is
more similar to inner speech than to talking. We next
asked whether inner speech might have the same effects
as talking. Using the same methods we used in Experiment 1, we substituted the Talking (aloud) condition
with an Inner Speech condition. Details of this experiment appear in an earlier report (Ford et al., 2001b).
5.2. Procedure
Subjects included in this experiment are described in
Table 1.
Following the presentation of probes during the silent
baseline condition, self-recorded hallucinatory statements, repeated for the 30 s recording were alternated
with the subjects repeating that same statement silently
to themselves for 30 s. This listen/inner speech sequence
was repeated seven times, once for each of seven different
statements and lasted seven minutes. The same random
mix of vowel, noise, and checkerboard probes continued
while subjects spoke aloud or silently.
5.3. Results
N1 amplitude effects are illustrated in the Fig. 5. In
control subjects, N1 during baseline was larger than
during inner speech [F(1,14)=9.64, P=0.008]. In
patients, N1 during baseline was not significantly larger
than inner speech, [F(1,14)=2.51, P=0.14]. An
ANOVA for N1 to the checkerboard did not approach
significance for group, condition, or groupcondition.
None of the Spearman correlations between the N1
effect (i.e., baseline minus inner speech) and SAPS
summary scores for hallucinations and delusions and
BPRS scores for hallucinatory behavior and unusual
thought content was significant.
5.4. Discussion
Using neurophysiological methods, we demonstrated
that inner speech reduces responsiveness of auditory
cortex in control subjects. Compared to baseline, N1 to
Fig. 4. Experiemtn 2. ERPs recorded at Cz to the speech sounds as they are being produced (talking, blue lines) and during playback of the same
sounds (red lines). As expected, auditory cortical responsiveness is dampened during talking compared to playback in the controls, but not in the
patients. ERPs were filtered with a 2–8 Hz band-pass filter.
J.M. Ford, D.H. Mathalon / Journal of Psychiatric Research 38 (2004) 37–46
43
Fig. 5. Experiment 3. On the left is a schematic of our hypothesis of a dysfunctional corollary discharge mechanism in schizophrenia, and its
extension to inner speech. The sentence in this figure was used in the experiment. On the right are the ERPs recorded from Cz to an acoustic probe
stimulus (noise) for controls and patients hearing the probes during a baseline condition of silence (baseline) and while repeating sentences silently to
themselves (inner speech). The red lines depict auditory cortical activity to the probes during the silent baseline, and the blue lines depict auditory
cortical activity to the probes during inner speech. In the controls, auditory cortical activity is relatively suppressed (blue) during inner speech
compared to the silent baseline (red); this is not true in the patients.
acoustic probes was reduced when control subjects
were repeating sentences silently to themselves. This
was not true in patients, for whom silent speech did
not reduce auditory cortical responsiveness perhaps
through poorly functioning corollary discharge (Feinberg, 1978). Corollary discharge may signal speech
reception areas that speech-related activations are selfgenerated, avoiding misperceptions that these thoughts
have an external source.
6. Experiment 4: EEG coherence during talking and
listening
6.1. Introduction
ERP evidence from the first three experiments suggests that auditory cortical responsiveness is reduced
during talking which we assume is due to a corollary
discharge from frontal speech producing areas to the
speech reception areas in the temporal lobe. However,
we have no direct evidence that the frontal lobe is
involved. To assess the role of frontal speech area
involvement, we calculated the degree of inter-relatedness between frontal and temporal lobes during talking
compared to listening, using EEG coherence algorithms.
Coherence is a frequency-dependent measure of the
degree of relatedness between EEG recorded over two
different brain areas. High coherence between two brain
areas indicates that their amplitudes at a given frequency and their associated phase angles are correlated
across time epochs (Lachaux et al., 1999). When coherence
is low, it indicates that across time epochs, the relationship
between power in the two signals and/or the relative phase
difference between them is inconsistent. Moreover, the
coherence measure does not allow specification of whether
the relationship between the two signals is stronger in terms
of relative phase or power (Lachaux et al., 1999).
Coherence can range from zero to one, and it does not
distinguish positive from negative correlations between
frequency amplitudes across time. Accordingly, it can
reflect either inhibition or excitation of connected areas
(Manganotti et al., 1998). Details of this study appear in
an earlier report (Ford et al., 2002).
44
J.M. Ford, D.H. Mathalon / Journal of Psychiatric Research 38 (2004) 37–46
6.2. Methods
Single trial EEG epochs were used from Experiment 1.
To translate data from the time domain to the frequency
domain, a Fast Fourier Transform (FFT) was calculated
on all points. Coherence was calculated for each frequency
band of interest (delta: 1–3 Hz; theta: 4–7 Hz; alpha: 8–12
Hz; beta: 13–20 Hz; gamma: 30–50 Hz). Coherence is the
spectral cross-correlation between two electrodes normalized by their power spectra (NeuroscanLabs, 1999).
Coherence was calculated between the following electrode pairs displayed in Fig. 6. The coherence values for
each electrode pair were averaged across the single trials.
The resulting average event-related coherence values for
each electrode pair were the dependent variables in the
analyses of variance (ANOVA).
6.3. Results
While the ANOVA revealed a main effect of
Condition [F(1,20)=26.28, P < 0.0001] indicating
greater coherence during talking than listening, the
Condition effect interacted with all the other variables
resulting in a 6-way interaction [GroupCondition
HemisphereFrontal RegionTemporal SiteFrequency
Band: F(6,320)=2.12, P < 0.05] which was parsed hierarchically to find a simple main effect of Condition,
proceeding only if the highest order interaction was significant in the intermediate ANOVA (from 4-way, to 3way, to 2-way). A 2-way GroupCondition interaction
was significant for theta [F(1,20)=4.62, P< 0.05] over the
left hemisphere between lateral frontal (F7) and posterior
temporal sites (P3). A similar GroupCondition interaction was significant for delta [F(1,20)=4.48, P< 0.05]
over the left hemisphere between lateral frontal (F7) and
posterior temporal sites (T5). Finally, at the single factor
level, a simple main effect of Condition was observed for
controls in both bands [Theta: F(1,9)=17.17, P=0.0025;
Delta: F(1,9)=5.05, P=0.05], but not for patients [Theta:
F(1,11)=3.28, P=0.10; Delta: F(1,11)=4.04, P=0.07].
Thus, for controls but not patients, coherence between lateral frontal and posterior temporal sites was greater during
Talking than Listening in the theta and delta bands.
Because the effects were stronger for theta than delta,
additional analyses focused on this frequency band. The
patient group was subdivided into hallucinators (rating
of 5, 6 or 7 on the Hallucinatory Behavior item on the
BPRS, n=7) and non-hallucinators (rating of 1 or 2 on
the Hallucinatory Behavior item on the BPRS, n=5).
We found a significant ConditionGroup interaction
for theta coherence [F(2,19)=4.14, P=0.03] in which
coherence was greater during Talking than Listening for
controls [F(1,9)=17.17, P=0.0025], tended to be greater
for non-hallucinators [F(1,4)=7.10, P=0.056], but was
Fig. 6. Experiment 4. Probability levels for t-tests showing greater fronto-temporal EEG coherence during talking than during listening superimposed on lateral view of right and left hemispheres for normal controls and patients with schizophrenia. Thicker lines indicate greater coherence
during talking than listening. In controls, coherence during talking was greater than during listening for all 20 of the electrode pairs. In patients, this
was true for only two of the pairs. Reprinted from Ford et al. (2002) with permission from the Society of Biological Psychiatry.
J.M. Ford, D.H. Mathalon / Journal of Psychiatric Research 38 (2004) 37–46
not for the hallucinators [F(1,6)=0.069, P=0.80]. Thus
the failure to increase theta coherence during talking in
schizophrenia is primarily observed in the hallucinators.
We compared theta coherence during Talking and Listening for each electrode pair. The resulting t-values and
probability levels are portrayed graphically in Fig. 6. In
controls, coherence during Talking was greater than during Listening for all 20 of the electrode pairs. In patients,
this was true for only two of the pairs.
6.4. Discussion
Using EEG coherence as a measure of functional connection between frontal and temporal brain areas, our
results corroborate other functional brain imaging reports
of a disconnection between frontal and temporal areas in
schizophrenia (e.g., Fletcher et al., 1999, Friston & Frith,
1995, Friston et al., 1995, Norman et al., 1997). The current coherence data are also consistent with the ERP data
from Experiments 1, 2, and 3, and augment them by suggesting that there is an interdependence between these areas
during talking, that is somewhat disrupted in patients,
especially those prone to auditory hallucinations. The
greater frontal-temporal coherence during talking than listening in controls may reflect the action of a corollary discharge from frontal brain structures preparing auditory
cortex for speech. It is important to note that because of the
nature of the coherence statistic, we cannot tell whether
high coherences reflect inhibition or excitation. That is, high
levels of activation in the frontal speech areas could be
related to high levels of excitation or inhibition in the temporal lobes. We do know however that 100 msec before
speech is initiated, the spontaneous firing rate of neurons in
the middle temporal gyrus diminishes (Creutzfeldt et al.,
1989b). Increased coherence during talking implies a continuous dialogue between neural systems responsible for
producing speech and those involved in perceiving its
effects, and this increase is not seen in patients with schizophrenia in the theta and delta bands. Furthermore, this
effect in the theta band was stronger in patients who hallucinated than those who did not. This would suggest that a
‘‘break’’ in the frontal-temporal circuit during the act of
overt speech, and perhaps covert speech, is associated with
the pathophysiology of auditory verbal hallucinations,
possibly because corollary discharge mechanisms normally
subserved by this circuitry are compromised. Whether the
connection between speech production and speech reception areas is subjectively experienced or whether it is an
unconscious automatic signal to the speech reception areas
remains an interesting and open question.
7. Summary of Experiments 1, 2, 3, and 4
Data from the first two experiments suggest that the
auditory cortex is dampened in responsiveness during
45
talking, confirming the primate (Muller-Preuss & Ploog,
1981) and intra-operative (Creutzfeldt et al., 1989a;
Creutzfeldt et al., 1989b) data of others, and extending
the MEG data of Curio and colleagues (Curio et al.,
2000) to the ERP methodology. In addition, the data from
the third experiment suggest that this dampening effect is
also seen during inner speech. Data from the fourth
experiment suggest that the suppression of the auditory
cortex responsiveness is due to a connection between the
frontal and temporal lobes. In each, we have shown evidence that normal effects of talking or inner speech do not
operate in patients with schizophrenia.
While we have chosen to interpret the data in terms of
the action of the corollary discharge mechanism, there
are other reasonable competing hypotheses. Among
these is the ‘‘line busy’’ hypothesis. According to this
view, talking, inner speech, or an on-going internal dialogue, might saturate the auditory cortex, making it
unresponsive to external probe stimuli, as if the ‘‘line’’
were ‘‘busy.’’ N1 to probes would be reduced if the line
were busy or if the corollary discharge was operating,
and with these data it is difficult to decide between these
two hypotheses. Alternatively, the corollary discharge
could be transmitted perfectly well from frontal lobe
speech production areas to temporal lobe speech reception areas in patients, but the temporal lobe structures are
dysfunctional and relatively unresponsive in schizophrenia, independent of hallucinations. This could be
referred to as the ‘‘nobody home’’ hypothesis. Another
explanation for our finding is that normal subjects but not
patients with schizophrenia may automatically increase
their auditory sensitivity when they are not talking in a
manner that can be suppressed by a corollary discharge.
In this paper we present an overview of a series of
studies performed to follow up on our initial observations of corollary discharge dysfunction in schizophrenia. The presentation is designed to focus on those
results that motivated each successive follow-up study
or analysis, rather than reiterate each individual study
in full, as these data have already been published. The
data presented here, while highlighting a consistent
theme, can certainly be subject to different interpretations, as noted above. Furthermore, the studies are all
limited by small sample size. Future studies will increase
sample size as well as expand the direction of the
enquiry. Another limitation of these studies is the limited
association detected between actual hallucinatory behavior and electrophysiological indicators of corollary discharge. While significant differences in theta coherence
emerged between patients classified as hallucinators and
non-hallucinators on the basis of BPRS scores in
Experiment 4, efforts to relate N1 measures to scores for
recent hallucinatory behavior derived from either BPRS
or the SAPS were unsuccessful. This lack of association
could be due to the normalizing effects of medication on
the symptom but not the mechanism that makes them
46
J.M. Ford, D.H. Mathalon / Journal of Psychiatric Research 38 (2004) 37–46
possible. Thus N1 measures may reflect the potential for
hallucinations rather than their current manifestation.
Future studies aim to test larger samples of patients
with and without histories of hallucinations and use
more comprehensive assessments of this symptom.
Acknowledgements
We acknowledge the help and support of our coauthors on the four papers summarized in this article:
Walton T. Roth, M.D., Susan Whitfield, William O.
Faustman, Ph.D., Sontine Kalba, and Theda Heinks.
We thank Mark Rothrock, M.D. for helping us recruit
patient volunteers to our studies, and we thank the
patients for their willingness to participant. We also
thank Margaret J. Rosenbloom for help preparing this
manuscript. This work was supported by grants from
National Institute of Mental Health (MH40052, MH
58262) and the Department of Veterans Affairs.
References
Andreasen NC. Scale for the assessment of positive symptoms. Iowa
City, IA: University of Iowa; 1984.
Blakemore S, Rees G, Frith C. How do we predict the consequences of our
actions? A functional imaging study. Neuropsychologia 1998;36:521–9.
Creutzfeldt O, Ojeman G, Lettich E. Neuronal activity in the human
lateral temporal lobe. I Responses to speech. Experimental Brain
Research 1989a;77:451–75.
Creutzfeldt O, Ojeman G, Lettich E. Neuronal activity in the human
lateral temporal lobe. II Responses to the subject’s own voice.
Experimental Brain Research 1989b;77:476–89.
Curio G, Neuloh G, Numminen J, Jousmaki V, Hari R. Speaking
modifies voice-evoked activity in the human auditory cortex.
Human Brain Mapping 2000;9:183–91.
Davis H, Zerlin S. Acoustic relations of the human vertex potential.
Journal of the Acoustic Society of America 1966;39:109–16.
Dierks T, Linden D, Jandl M, Formisano E, Goebel R, Lanfermann
H, Singer W. Activation of Heschl’s Gyrus during auditory hallucinations. Neuron 1999;22:615–21.
Feinberg I. Efference copy and corollary discharge: implications for
thinking and its disorders. Schizophrenia Bulletin 1978;4:636–40.
Feinberg I, Guazzelli M. Schizophrenia—a disorder of the corollary
discharge systems that integrate the motor systems of thought with
the sensory systems of consciousness. British Journal of Psychiatry
1999;174:196–204.
First MB, Spitzer RL, Gibbon M, Williams JBW. Structured clinical
interview for DSM-IV axis I disorders. New York, NY: Biometrics
Research Department, New York State Psychiatric Institute; 1995.
Fletcher P, McKenna PJ, Friston KJ, Frith CD, Dolan RJ. Abnormal
cingulate modulation of fronto-temporal connectivity in schizophrenia. Neuroimage 1999;9:337–42.
Ford JM, Mathalon DH, Heinks T, Kalba S, Roth WT. Neurophysiological evidence of corollary discharge dysfunction in schizophrenia. American Journal of Psychiatry 2001a;158:2069–71.
Ford JM, Mathalon DH, Kalba S, Whitfield S, Faustman WO, Roth
WT. Cortical responsiveness during inner speech in schizophrenia:
an event-related brain potential study. American Journal of Psychiatry 2001b;158:1914–6.
Ford JM, Mathalon DH, Kalba S, Whitfield S, Faustman WO, Roth
WT. Cortical responsiveness during talking and listening in schizo-
phrenia: an event-related brain potential study. Biological Psychiatry 2001c;50:540–9.
Ford JM, Mathalon DH, Whitfield S, Faustman WO, Roth WT.
Reduced communication between frontal and temporal lobes during
talking in schizophrenia. Biological Psychiatry 2002;21:485–92.
Friston KJ, Frith CD. Schizophrenia: a disconnection syndrome?
Clinical Neurosciences 1995;3:89–97.
Fristonm KJ, Holmes A, Poline J-B, Price CJ, Frith CD. Detecting
activations in PET and fMRI: Levels of inference and power. Neuroimage 1995;4:223–36.
Frith CD, Done DJ. Experiences of alien control in schizophrenia
reflect a disorder in the central monitoring of action. Psychological
Medicine 1989;19:359–63.
Frith CD, Friston K, Liddle PF, Frackowiak RSJ. Willed action and the
prefrontal cortex in man—a study with PET. Proceedings of the Royal
Society of London Series B—Biological Sciences 1991;244:241–6.
Han S, Liu W, Yund E, Woods D. Interactions between spatial
attention and global/local feature selection: an ERP study. Neuroreport 2000;11:2753–8.
Hedlund JL, Vieweg BW. The Brief Psychiatric Rating Scale (BPRS):
a comprehensive review. Journal of Operational Psychiatry 1980;11:
48–64.
Hillyard SA, Hink RF, Schwent VL, Picton TW. Electrical signs of
selective attention in the human brain. Science 1973;182:177–80.
Jackson JH. Selected writings of John Hughlings Jackson. In: Taylor
J. New York: Basic Books; 1958.
Lachaux J, Rodriguez E, Martinerie J, Varela F. Measuring phase
synchrony in brain signals. Human Brain Mapping 1999;8:194–208.
Manganotti P, Gerloff C, Toro C, Katsuta H, Sadato N, Zhuang P,
Leocani L, Hallett M. Task-related coherence and task-related
spectral power changes during sequential finger movements. Electroencephalography and clinical Neurophysiology 1998;109:50–62.
Martinez A, Anllo-Vento L, Sereno M, Frank L, Buxton R, Dubowitz
D, Wong E, Hinrichs H, Heinze H, Hillyard S. Involvement of
striate and extrastriate visual cortical areas in spatial attention.
Nature Neuroscience 1999;2:364–9.
Muller-Preuss P, Ploog D. Inhibition of auditory cortical neurons
during phonation. Brain Research 1981;215:61–76.
NeuroscanLabs. Edit 4.1 Offline Analysis of Acquired Data: Coherence. SCANTM User Guide Volume II. Sterling, VA: Neurosoft,
Inc; 1999.
Norman RMG, Malla AK, Williamson PC, Morrison-Stewart SL,
Helmes E, Cortese L. EEG coherence and syndromes in schizophrenia. British Journal of Psychiatry 1997;170:411–5.
Numminen J, Salmelin R, Hari R. Subject’s own speach reduces reactivity
of the human auditory cortex. Neuroscience Letters 1999;265:119–22.
Overall JE, Hollister L, Pichot R. Major psychiatric disorders. A fourdimensional model. Archives of General Psychiatry 1967;16:146–51.
Papanicolaou A, Wilson G, Busch C, DeRego P, Orr C, Davis I,
Eisenberg H. Hemispheric asymmetries in phonological processing
assessed with probe evoked magnetic fields. International Journal of
Neurosciences 1988;39:275–81.
Paus T, Perry DW, Zatorre RJ, Worsley KJ, Evans AC. Modulation
of cerebral blood flow in the human auditory cortex during speech:
role of motor-to-sensory discharges. European Journal of Neuroscience 1996;8:2236–46.
Shergill S, Brammer M, Williams S, Murray R, McGuire P. Mapping
auditory hallucinations in schizophrenia using functional magnetic
resonance imaging. Archives of General Psychiatry 2000;57:1033–8.
Sperry RW. Neural basis of the spontaneous optokinetic response
produced by visual inversion. Journal of Comparative and Physiological Psychology 1950;43:482–9.
Von Holst E, Mittelstaedt H. Das reafferenzprinzip. Naturwisshenschaften 1950;37:464–76.
Warburton E, Wise RJS, Price CJ, Weiller C, Hadar U, Ramsay S,
Frackowiak RSJ. Noun and verb retrieval by normal subjects.
studies with PET. Brain 1996;119:159–79.